Identification of Small Molecules That Can Activate the Nurr1 Function.

To identify small molecules that can activate Nurr1, we established efficient high-throughput assay systems based on our previous study showing that Nurr1 prominently activates the tyrosine hydroxylase (TH) gene promoter function (20). We used the reporter construct p4xNL3-Luc (containing four copies of the NGFI-B responsive element (NBRE)-like motif (NL3), identified in the TH upstream promoter positioned immediately proximal to the TATA box and linked to the luciferase gene) along with pCMV-Nurr1 construct (full-length Nurr1-expression vector) in human neuroblastoma SK-N-BE(2)C cells, followed by treatment with each compound for 24 h followed by measurement of luciferase activity. Using these cell-based assay systems, we screened a chemical library composed of 960 FDA-approved drugs (MicroSource Discovery Systems) and successfully identified three hit compounds (i.e., two antimalaria drugs amodiaquine (AQ) and chloroquine (CQ) and a pain-relieving drug (glafenine). Remarkably, all three compounds share an identical chemical scaffold, 4-amino-7-chloroquinoline, suggesting a SAR associated with this motif (Fig. 1A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Identification of AQ and CQ as Nurr1 activators. (A) Chemical structures of three hit compounds that activate Nurr1’s transcriptional function. Notably, all compounds contain an identical scaffold, 4-amino-7-chloroquinoline (highlighted by red color). (B) AQ, CQ, and glafenine increase the transcriptional activity of Nurr1-based reporter constructs: full-length Nurr1-dependent (fNurr; Left) and Nurr1 LBD-dependent (Right) transcriptional activities. (C) Effect of Nurr1-specific siRNA on Nurr1 LBD’s transactivation activity. A knockdown of Nurr1 expression by treatment with Nurr1-specific siRNA reduced the reporter gene activity of the Nurr1 LBD construct. (D) Effect of SRC proteins on AQ- and CQ-induced Nurr1 transactivation. AQ and CQ enhanced Nurr1’s transcriptional function in the presence of SRC-1/SRC-3 overexpression. The fold induction was derived by comparing each luciferase activity to basal level obtained by non-SRC transfection. (E) Target selectivity of AQ/CQ for LBDs of various NRs. A total of 30 μM AQ and 100 μM CQ robustly activates LBD function of Nurr1, but not other NRs tested here, indicating a high specificity. The positive reactivity of these NR constructs was confirmed by the known activators (2 nM dexamethasone, 20 nM retinoic acid, 2 μM GW3965, 50 nM GW7647, and 5 nM GW1929 for glucocorticoid receptor (GR), retinoid X receptor-α (RXRα), liver X receptor-α (LXRα), peroxisome proliferator-activated receptor-α (PPARα), and PPARγ, respectively). The basal level of transcriptional activity was normalized to 1. Bars represent means ± SEM from three independent experiments.

In our original assays using the reporter construct p4xNL3-Luc containing four copies of the NBRE-like NL3 motif residing in the TH promoter (20), AQ and CQ activated luciferase reporter activity up to approximately 3-fold in a dose-dependent manner (Fig. 1B and SI Appendix, Fig. S1 A and B). In the present study we focused on AQ and CQ because glafenine showed a weaker activity (∼1.5-fold). To investigate whether AQ and CQ activate Nurr1 function through its LBD or DNA binding domain (DBD), we generated additional reporter constructs in which the yeast transcription factor GAL4’s DBD is fused to either Nurr1 LBD or DBD [pGAL-Nurr(LBD) and pGAL-Nurr(DBD)]. Notably, AQ and CQ stimulated Nurr1-dependent transcriptional activity through its LBD, whereas no response was observed when GAL4 or GAL4-Nurr1 DBD was used (Fig. 1B, Right). AQ and CQ induced Nurr1 LBD-based reporter activity up to ∼15- and 10-fold with an EC₅₀ of ∼20 and 50 μM, respectively (Fig. 1B and SI Appendix, Fig. S1 C and D).

To test if reporter gene activation by AQ and CQ is through Nurr1, we treated cells with Nurr1-specific siRNA and confirmed decreased levels of Nurr1. Nurr1 siRNAs, but not scrambled RNAs, reduced Nurr1 protein expression as well as AQ/CQ-induced luciferase activation by more than 80%, strongly suggesting that transcriptional activation by AQ and CQ is indeed through modulation of Nurr1 function (Fig. 1C and SI Appendix, Figs. S2 and S3). The recruitment of transcriptional coregulators such as steroid receptor coactivator (SRC) is a crucial mechanism of nuclear receptors (NRs) (21). To test if AQ/CQ modulates Nurr1’s interaction with SRC, a coimmunoprecipitation assay was performed. We found that Nurr1 weakly interacts with SRC-1 and SRC-3 in SK-N-BE(2)C cells and that these interactions were significantly enhanced by AQ and CQ (SI Appendix, Fig. S4). We next tested the effects of AQ and CQ on Nurr1 transcriptional function following SRC-1 or SRC-3 overexpression. Whereas SRC-1/SRC-3 overexpression itself did not greatly influence Nurr1 function, AQ and CQ further enhanced Nurr1’s transcriptional function in the presence of SRC-1/SRC-3 overexpression (Fig. 1D). Taken together, our data suggest that AQ and CQ induce Nurr1’s transactivation function through its Nurr1 LBD by facilitating the recruitment of coactivators like SRC-1/SRC-3.

We next tested if AQ and CQ are able to nonspecifically activate the LBD function of other NRs. As shown in Fig. 1E, AQ and CQ were unable to induce the transcriptional activity of any of these constructs, showing high selectivity. We also tested if various quinoline compounds have similar Nurr1-activating function. Remarkably, none of these compounds showed any detectable transactivation function in a wide range of concentrations (SI Appendix, Fig. S5). Because only AQ and CQ contain the 4-amino-7-chloroquinoline entity, these results further support its SAR.

AQ/CQ Physically Binds to the Nurr1.

Our findings are surprising because Nurr1 lacks a “classical” binding pocket due to the presence of bulky hydrophobic side chain residues and is thought to be a ligand-independent and constitutively active NR (22, 23). Thus, we further tested whether AQ and CQ physically interact with Nurr1’s LBD. Toward this goal, we purified the Nurr1 LBD polypeptide (amino acids 328–598) and examined its binding with AQ. First, we analyzed their physical binding using the Biacore S51 SPR sensor, which has a higher sensitivity and improved fluidics, enabling the monitoring of small signal changes derived from binding of compounds to proteins. As shown in SI Appendix, Fig. S6 A and B, AQ specifically bound to Nurr1-LBD in a dose-dependent manner but not to the retinoid X receptor (RXR)-LBD. Next, we performed fluorescence quenching analysis. The Nurr1-LBD displayed maximal fluorescence at 336 nm, whereas AQ itself had no fluorescence at this wavelength. When the Nurr1-LBD was incubated with increasing amounts of AQ, fluorescence intensity gradually decreased (SI Appendix, Fig. S6C). In contrast, RXR-LBD’s fluorescence emission was quenched by 9-cis retinoic acid but not by AQ over a wide range of concentrations (SI Appendix, Fig. S6 D and E). In addition, we performed a radioligand binding assay using [³H]-CQ. [³H]-CQ showed saturable binding to Nurr1-LBD with a dissociation constant (K_d) of 0.27 μM and a maximal binding capacity (B_max) of 13.9 μM (Fig. 2A). Furthermore, competition-binding assay showed that unlabeled AQ/CQ can compete for binding with [³H]-CQ with K_i values of 246 and 88 nM for AQ and CQ, respectively (Fig. 2B). In contrast, unlabeled primaquine, which could not enhance Nurr1’s transcriptional function (SI Appendix, Fig. S5), was unable to compete. To further investigate the molecular interaction between AQ and the Nurr1 LBD, we examined the chemical shift perturbations (CSPs) on 2D ¹H-¹⁵N heteronuclear single quantum correlation (HSQC) spectra of the ¹⁵N-labeled Nurr1-LBD in the presence of AQ. A number of LBD residues showed notable CSP upon addition of AQ (Fig. 2 C and D and SI Appendix, Fig. S7). The overlay of free and ligand-bound spectra of the Nurr1 LBD showed that perturbed residues are mainly located in the helix α2 region (H402, I403, Q404, Q405, D408, and L409) and some are in α11 (V468, Y575, and D580) (SI Appendix, Fig. S7B). Thus, it appears that AQ binding at α2 interacts with residues Q474 and D408 (within the vicinity of L409). However, as AQ at this binding pose is unlikely to form any direct interactions with C-terminal helix 11/12 residues, CSPs observed in Y575 and D580 might be attributable to either conformational changes or an allosteric effect upon AQ binding. To assess the functional contribution, we mutated the perturbed residues in H2 and H11 to alanine and tested their transcriptional activities. As shown in Fig. 2E, mutants I403A, L409A, Y575A, and D580A exhibited profound reduction of transcriptional activity, whereas other mutations largely retained the activity. This functional analysis supports that I403, L409, Y575, and D580, identified from nuclear magnetic resonance (NMR) experiments, are critical residues involved in the physical interaction of AQ and activator recognition. Together, these data strongly suggest that AQ activates Nurr1 function via direct and specific binding to the Nurr1 LBD.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Interaction of AQ/CQ with the Nurr1-LBD protein. (A) Nurr1-LBD protein was incubated with increasing concentrations (3.9, 7.8, 15.5, 31, 62.5, 125, 250, 500, and 1,000 nM) of [³H]-CQ. The Inset indicates Scatchard analysis of the specific binding. (B) Competition of AQ, CQ, and primaquine (PQ) for binding of [³H]-CQ to Nurr1-LBD. Increasing concentrations of unlabeled AQ, CQ, or PQ were incubated with 500,nM [³H]-CQ and Nurr1-LBD. (C) Molecular interaction of the Nurr1-LBD and AQ by NMR titration experiments using uniformly ¹⁵N-labeled Nurr1-LBD. The 2D ¹H-¹⁵N TROSY-HSQC spectra of Nurr1-LBD were recorded on a Bruker Avance 700 spectrometer at 298K in the absence (red) and presence of AQ at molar ratios of 1–1 (magenta), 1–2 (black), and 1–5 (blue). Expanded sections of overlaid 2D ¹H-¹⁵N TROSY-HSQC spectra show concentration-dependent chemical shift perturbations upon AQ binding. Amino acids showing chemical shift perturbations with increasing concentration of AQ are indicated by arrows. Disappeared resonance of I403 by addition of AQ is marked as rectangular box. (D) Mapping of the interaction sites between Nurr1-LBD and AQ. (Left) Surface mapping of AQ binding site and interaction residues on the crystal structure of Nurr1-LBD based on 2D ¹H-¹⁵N HSQC titration data. Perturbed amino acid residues were displayed according to their chemical shift perturbation values: red (Δδ > 0.1), blue (0.08 < Δδ < 0.1), and green (disappeared), respectively. (Right) Expanded view of potential binding pocket for amodiaquine on the Nurr1-LBD. Perturbed amino acid residues were displayed by the same manner on the Right. (E) Functional effects of mutations in the potential AQ binding residues on Nurr1’s transcriptional activity. Wild-type and mutant constructs were tested by transient transfection assay with or without AQ. The mutations at I403, L409, Y575, or D580 significantly reduced both basal transcriptional activity and its activation by AQ.

AQ/CQ Promotes mDA Neurogenesis.

We next sought to test whether these compounds exhibit biological effects on mDA neurons in more physiological contexts. First, we tested if they can increase the generation of TH⁺ neurons and/or expression of the mDA-specific genes during in vitro differentiation of neural stem cells. Indeed, we found that AQ enhanced Nurr1’s function both in terms of the number of TH⁺ neurons generated from neural stem cells (Fig. 3 A and B) and mRNA expression levels of the TH, dopamine transporter (DAT), vesicular monoamine transporter (VMAT), and aromatic amino acid decarboxylase (AADC) genes (Fig. 3C). Next, we performed qRT-PCR in the absence or presence of siRNA to Nurr1 to test whether induction of gene expression by AQ/CQ is abolished by siRNA. As shown in Fig. 2C, siRNA to Nurr1 abrogated gene expression of TH, DAT, VMAT2, and AADC induced by AQ/CQ (Fig. 3C). In vivo chromatin immunoprecipitation (ChIP) assay clearly showed that Nurr1 is recruited to both NL1 and NL3 sites of the TH promoter in AQ- and CQ-treated cells (20), suggesting that Nurr1 is directly involved in AQ- and CQ-mediated induction of TH gene expression (Fig. 3D).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Functional effects of AQ and CQ. (A and B) AQ stimulated the generation and gene expression of DA neurons from neural progenitors isolated from E14.5 rat cortex in a dose-dependent manner. Differentiation was induced by withdrawal of bFGF and AQ was added for 2 h during the differentiation. Immunocytochemical analyses for TH (A) and yields of TH⁺/DAPI cells (B) for each treatment group were obtained following in vitro differentiation for 3 d and 9 d. (C) Real-time PCR analysis shows that AQ treatment enhances expression of the mDA-specific genes TH, dopamine transporter (DAT), vesicular monoamine transporter (VMAT), and aromatic amino acid decarboxylase (AADC) during in vitro differentiation of neural stem cells at 9 d. (D) Rat PC12 cells were treated with 20 μM AQ or 70 μM CQ. ChIP assay shows AQ- or CQ-dependent Nurr1 recruitment to the TH promoter. *P < 0.05 and **P < 0.005 versus untreated. Results are expressed as the average of three independent experiments. Error bars represent SDs.

Neuroprotective and Inflammation-Modulating Effects of AQ/CQ.

Next, we tested if AQ and CQ prevent neurotoxin (6-OHDA)-induced death in primary DA neurons and rat PC12 cells. AQ and CQ significantly inhibit 6-OHDA–induced cell death in primary DA cells as examined by the number of TH⁺ neurons (Fig. 4A) and DA uptake (Fig. 4B). The neuroprotective effect of AQ and CQ was also observed in rat PC12 cells (Fig. 4C). Because Nurr1 has opposite transrepression activity in microglia and astrocytes (12), we also analyzed the effect of AQ on expression of proinflammatory cytokine genes in primary microglia derived from P1 rat brains. When these cells were treated with the inflammation-inducing lipopolysaccharide (LPS; 10 ng/mL) for 8 h, expression of proinflammation genes tested [interleukin-1β (IL-1β), interleukin-6, tumor necrosis factor-α (TNF-α), and inducible nitric oxide synthase (iNOS)] was dramatically induced (Fig. 4D). Remarkably, AQ treatment prominently reduced the expression of all these genes in a dose-dependent manner (>10-fold). AQ and CQ showed very similar effects for repressing these cytokine genes in both primary microglia and BV-2 microglial cell line (SI Appendix, Fig. S8 A and B). Taken together, our data show that AQ and CQ are able to enhance the contrasting dual roles of Nurr1: (i) they increase Nurr1’s transactivation of mDA-specific genes in mDA neurons, and (ii) they also further enhance Nurr1’s transrepression of proinflammatory cytokine gene expression in microglia.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Neuroprotective effects of AQ and CQ. (A–C) Primary cultures of rat mesencephalic DA neurons were treated with 20 μM 6-OHDA for 24 h in the presence or absence of 5 μM AQ and 20 μM CQ. (A) The number of TH⁺ neurons and (B) the rate of [³H]DA uptake were measured. (C) Cell survival was measured by the MTT reduction assay in PC12 cells treated with 6-OHDA alone or in combination with AQ. Values from each treatment expressed as a percentage of untreated control for the MTT assay. (D) AQ suppresses LPS-induced expression of proinflammatory cytokines. Primary microglia from P1 rat brains were treated with 10 ng/mL LPS for 4 h in the presence or absence of AQ (10, 15, and 20 μM). Levels of mRNA expression were analyzed by quantitative real-time PCR and normalized with GAPDH. This experiment has been repeated three times in triplicate using independently prepared RNAs. Each bar represents means ± SEM of n = 4–5. *P < 0.05, **P < 0.01,***P < 0.001, compared with the LPS-only–treated group.

AQ Improves Behavioral Deficits in a 6-OHDA–Lesioned Rat Model of PD.

The above findings prompted us to test whether these compounds can ameliorate motor behavior deficits in a PD animal model. Toward this goal, we tested the effects of AQ administration in an animal model of PD, 6-OHDA–lesioned rats. Rats with unilateral intrastriatal 6-OHDA lesion were administered with saline or AQ for two weeks, starting from 1 d before intrastriatal 6-OHDA lesion (Fig. 5A). As expected, amphetamine administration produced rotation behavior toward the lesion side in control rats, when measured at 4 and 6 wk post–6-OHDA lesion, indicating unilateral damage to the right striatum (Fig. 5 A and B). Remarkably, AQ administration significantly ameliorated 6-OHDA–induced rotation behavior at 4 and 6 wk post–6-OHDA lesion (P < 0.06 and P < 0.0003, respectively). To address if AQ treatment triggers abnormal involuntary movements (AIMs), well-validated dyskinesia-like behaviors (24), we measured AIM scores such as axial, limb, locomotive, and orolingual dyskinesias. In l-DOPA–treated rats (8 mg/kg for 2 wk before scoring), as expected, the animals showed dramatically increased AIM scores [F(2,71) = 33, P < 0.0001], classified as the axial, limb, locomotive, and orolingual dyskinesias (Fig. 5C). In contrast, both saline- and AQ-treated 6-OHDA rats did not show any detectable AIM behavior.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Effects of AQ treatments on a 6-OHDA–lesioned rat model of PD. (A) Schematic representation of the administration of AQ to 6-OHDA–lesioned rats. Unilateral striatal 6-OHDA–lesioned rats were treated with AQ or saline for 2 wk, starting from 1 d before lesioning. The gray shade indicates l-DOPA treatment for 2 wk as control of AIMs test. (B) Amphetamine-induced rotational test was performed at 4 and 6 wk post–6-OHDA lesion. (C) 6-OHDA–lesioned rats treated with l-DOPA, but not with AQ, exhibited severe side effects, as measured by AIMs scores. Four types of AIMs were monitored (axial, forelimb, orolingual, and locomotor) at 2 and 6 wk postlesioning. Each AIM behavior was monitored and scored on a 0–4 scale and summed. Bars represent the mean + SEM (^#P < 0.06, *P < 0.01, ***P < 0.0001).

Immunohistochemistry of the substantia nigra (SN) and the striatum (STR) of these animals showed that AQ-treated rats retained a significant number of TH⁺ cells in the lesion side SN, whereas saline-treated control rats lost the great majority of TH⁺ cells (Fig. 6A). Similarly, abundant TH⁺ fibers were spared in the STR of AQ-treated brains, whereas they were mostly lost in the STR of control brains. To quantitatively analyze these data, TH⁺ cells in the SN were stereologically counted in a blind manner. As shown in SI Appendix, Fig. S9, the number of TH⁺ neurons in the lesion side of AQ-treated rats was ∼60% of those from the intact side at 6 wk post–6-OHDA lesion (P < 0.0003), whereas it remained less than 20% in saline-treated animals. Importantly, these spared TH⁺ cells coexpressed other DA markers such as FoxA2 and AADC (SI Appendix, Fig. S10). We also examined microglial activation by immunohistochemistry of the microglial marker ionized calcium binding adaptor molecule-1 (Iba-1). As shown in Fig. 6 B–E, microglial activation was prominently observed in the ipsilateral SN (Fig. 6B) and STR regions (Fig. 6C) of 6-OHDA–lesioned rats. In contrast, following AQ treatment, the numbers of Iba-1⁺ microglia in the ipsilateral SN (Fig. 6 B and D) and STR regions (Fig. 6 C and E) decreased to the levels of the contralateral sides, indicating a robust suppression of neuroinflammation by Nurr1 activation (12).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Immunocytochemical analysis of 6-OHDA–lesioned rat administrated by AQ. (A) TH⁺ fibers and neurons were plentiful in the STR and SN of normal site, whereas their marked depletion was observed in the right STR and SN by 6-OHDA lesioning. In contrast, abundant TH-immunopositive cells were spared not only in the striatum but also in the SN in the AQ-treated group. [Scale bar, 200 μm (black), 100 μm (white), 20 μm (red).] (B–E) AQ reduces microglial activation in 6-OHDA–injected rat brains. Brain sections including the SN (B and D) and STR (C and E) regions were immunostained with anti–Iba-1 antibody, and Iba-1⁺ cells were counted in both lesion and intact sides. Data represent the mean ± SEM (**P < 0.01, ***P < 0.001). (Scale bar, 200 μm.)

In summary, using our high throughput cell-based assays, we identified small molecules, AQ and CQ, which can activate Nurr1 through its LBD and significantly improve motor impairments in PD animal model without dyskinesia-like side effects. Our experimental evidence supports that these compounds enhanced the contrasting dual functions of Nurr1, activation of mDA neuron-specific function (e.g., TH expression) and repression of microglial activation and neurotoxic cytokine gene expression, leading to significant neuroprotective and/or neurorestorative effects in a PD animal model. Notably, based on selective binding to neuromelanin to 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) metabolite (25), a previous study demonstrated that administration of CQ to monkeys protected them from MPTP-induced parkinsonian motor abnormalities (26). Our findings corroborate CQ’s neuroprotective effect and further suggest the involvement of alternate mechanism related to Nurr1 activation. Together with this interesting primate study, our results provide a preclinical “proof of concept” that Nurr1 could serve as a valid target for further development of mechanism-based, disease-modifying therapeutics of PD.